dlmarion opened a new issue, #4420: URL: https://github.com/apache/accumulo/issues/4420
The documentation for the `srv:lock` column is: ``` srv:lock [] tservers/127.0.0.1:9997/zlock-0000000001\$13fe86cd27101e5 - This is the lock information for the tablet holding the present lock. This information is checked against zookeeper whenever this is updated, which prevents a metadata update from a tablet server that no longer holds its lock. ``` In `main`, the `srv:lock` column is set via `TabletMutator.putZooLock` in the client code, and it's checked in `MetadataConstraints.check` on the server side. `MetadataConstraints` checks that the lock value set in the Mutation is actually a valid lock in ZooKeeper. The ZooKeeper lock is used for validation instead of the location as a TabletServer could be restarted on the same host and end up with the same address, but the lock would not be the same. It's possible that TabletServer A could send a mutation to TabletServer B to update the metadata table for a Tablet that TabletServer A is hosting, and then TabletServer A could fail right after sending the mutation. In `main`, `TabletMutator.putZooLock` is called from: ``` ManagerMetadataUtil.addNewTablet ManagerMetadataUtil.replaceDatafiles ManagerMetadataUtil.updateTabletDataFile MetadataTableUtil.addTablet MetadataTableUtil.removeScanFiles MetadataTableUtil.removeUnusedWALEntries MetadataTableUtil.updateTabletCompactID MetadataTableUtil.updateTabletDataFile MetadataTableUtil.updateTabletFlushID MetadataTableUtil.updateTabletVolumes ``` In `elasticity` the `putZooLock` method moved to `TabletUpdates` and is called from: ``` TabletGroupWatcher.replaceVolumes PopulateMetadata.writeSplitsToMetadataTable MetadataTableUtil.removeScanFiles MetadataTableUtil.removeUnusedWALEntries MetadataTableUtil.updateTabletDataFile Tablet.flush Tablet.updateTabletDataFile ``` The differences in these lists makes me question whether if we missed something when updating the code. I'm also wondering in which cases the ServiceLock should *not* be checked when updating that metadata and root tables - should it always be checked? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: notifications-unsubscr...@accumulo.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
